DXA based radiomics scoring: a novel approach for fragility fracture risk assessment in elderly type 2 diabetes patients.

This study develops a DXA-based radiomic model to predict fracture risk in elderly T2DM patients. The model, validated with AUCs up to 0.774, 0.759, and 0.739 for vertebral, hip, and major osteoporotic fractures respectively, offers a potential tool for personalized fracture risk assessment and management in clinical settings.

Purpose: To develop and validate a novel radiomic model based on dual-energy X-ray absorptiometry (DXA) lumbar spine and hip images for predicting fracture risk in elderly type 2 diabetes mellitus (T2DM) patients.

Methods: A training cohort of 294 elderly T2DM patients (October 2019 to August 2022) and a validation cohort from the ABCD cohort were utilized. Radiomics features were extracted from baseline DXA images, and a radiomics model was constructed using Cox proportional hazards regression. A user-friendly radiomics score (Radscore) was derived for practical application.

Results: In the derivation set, 105 (35.7%) patients had radiographically confirmed vertebral fractures. In the validation set with a median 5-year follow-up, 3.2% experienced incident clinical vertebral fracture, 2.4% incident hip fracture, and 7.0% incident major osteoportic fractures (MOFs). Thirteen features were selected to develop the radiomics model. The AUC for predicting radiographically confirmed vertebral fracture in the derivation set was 0.830 (95% CI: 0.783-0.878). In the validation set, AUCs for predicting incident clinical vertebral fractures, hip fractures, and MOFs were 0.774 (95% CI: 0.686-0.862), 0.759 (95% CI: 0.645-0.873), and 0.739 (95% CI: 0.676-0.803), respectively. The Radscore exhibited weak correlations with traditional fracture assessment tools but independently with the diabetes-specific phenotype. It effectively stratified T2DM patients into risk groups with significantly differing fracture incidences.

Conclusions: This DXA-based Radscore presented in this study offers a robust approach to stratify fracture risk in elderly patients with T2DM. Its potential integration into clinical settings could facilitate personalized management strategies aimed at reducing fracture incidence.